As marcas lexicais da discriminação étnico-racial entre negros e brancos: um estudo da linguística de corpus

• Ano de defesa: 2020
• Autor(a) principal: Souza, Rafael Webster Ferreira de
• Orientador(a): Sardinha, Antonio Paulo Berber
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica de São Paulo
• Programa de Pós-Graduação: Programa de Estudos Pós-Graduados em Linguística Aplicada e Estudos da Linguagem
• Departamento: Faculdade de Filosofia, Comunicação, Letras e Artes
• País: Brasil
• Palavras-chave em Português: Discriminação racial; Negros; Brancos; Linguística de corpus; Grupos étnicos; Linguagem e línguas - Variação
• Palavras-chave em Inglês: Race discrimination; Blacks; Whites; Corpora (Linguistics); Ethnic groups; Language and languages - Variation
• Área do conhecimento CNPq: CNPQ::LINGUISTICA, LETRAS E ARTES::LINGUISTICA::LINGUISTICA APLICADA
• Link de acesso: https://repositorio.pucsp.br/jspui/handle/handle/24088

Language serves as a means of social indexing, varying between groups of speakers and allowing social groups to distinguish between insiders and outsiders. In other words, it allows us to distinguish between individuals who belong and do not belong to a particular group (HALL, 2016; BAKER, 2010). The overall purpose of this study is threefold: (a) to identify the subject matters, themes, or discourses produced by black and white speakers of Brazilian Portuguese, (b) to determine which of these subject matters, themes or discourses are similar or different among black and white speakers, and (c) to analyze whether the ethnic-racial component of its speakers is likely to be predicted by the lexical variables used in the texts. These goals were pursued from a Corpus Linguistics perspective (HUNSTON; ESIMAJE, 2019; BAKER; EGBERT, 2016; BIBER, 2009; BERBER SARDINHA, 2004). Therefore, a corpus has been collected to represent speaking situations in which black and white speakers are involved in Brazilian Portuguese. The corpus was called ‘Corpus of Ethnic-Racial Registers’, whose name in Portuguese is Corpus de Registros Étnico-Raciais (CRER), comprising 788 texts representing several samples of language in use from different registers (namely popular songs, vlogs, and oral life histories) in Brazilian Portuguese. The texts for each register (and subregisters) in the corpus were balanced for black and white speakers. Once the corpus was fully built, the sample files were morphosyntactically tagged and the words tagged as nouns, adjectives, verbs and adverbs were selected for each text. The counts for those words were then normalized by 1,000 words. The resulting lexis variables were analyzed using two methods: Lexical Multidimensional Analysis (BERBER SARDINHA, 2014) and Discriminant Functional Analysis (BERBER SARDINHA; VEIRANO PINTO, 2014, 2019; NORRIS, 2015). For the former method, dimensions of lexical variation were statistically identified and communicatively interpreted to find the subject matters, themes or discourses. The statistical comparisons showed a high degree of difference across the registers for the dimensions, indicating that register is a powerful predictor of variation for language use, which corroborates previous multi-dimensional studies (BIBER, 1988; BERBER SARDINHA; VEIRANO PINTO, 2014, 2019). At the same time, the dimensions failed to reveal variation between black and white speakers. For the latter method, the selected lexis from the texts was entered in a Discriminant Functional Analysis returning sets of words considered predictors of each ethnic group. The statistical comparisons showed a high degree of accurate prediction for both ethnic groups (88,13% for blacks, and 72,29% for whites). Additionally, the predicting words of each group were interpreted in connection with the underlying subject matters, themes and discourses, which surprisingly revealed subtle, yet undeniable traces of discrimination, racism and bias toward blacks. In short, both methods showed different outcomes; while the dimensions of lexical variation did not reveal differences between black and white speakers, the Discriminant Analysis pointed out clear differences for both groups. Thus, we were able to conclude that the speakers’ ethnic group can be identified through the lexis employed in their subject matters, themes and discourses and, unexpectedly, that there is hidden racial discrimination and prejudice in the topics black and white Brazilians talk about every day